CN115063302B - Effective removing method for spiced salt noise of fingerprint image - Google Patents
Effective removing method for spiced salt noise of fingerprint image Download PDFInfo
- Publication number
- CN115063302B CN115063302B CN202210501241.9A CN202210501241A CN115063302B CN 115063302 B CN115063302 B CN 115063302B CN 202210501241 A CN202210501241 A CN 202210501241A CN 115063302 B CN115063302 B CN 115063302B
- Authority
- CN
- China
- Prior art keywords
- pixel
- window
- noise
- value
- fingerprint image
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 150000003839 salts Chemical class 0.000 title claims abstract description 30
- 238000000034 method Methods 0.000 title claims abstract description 28
- 239000006002 Pepper Substances 0.000 claims abstract description 37
- 235000002566 Capsicum Nutrition 0.000 claims abstract description 28
- 241000722363 Piper Species 0.000 claims abstract description 28
- 235000016761 Piper aduncum Nutrition 0.000 claims abstract description 28
- 235000017804 Piper guineense Nutrition 0.000 claims abstract description 28
- 235000008184 Piper nigrum Nutrition 0.000 claims abstract description 28
- 230000000694 effects Effects 0.000 claims abstract description 20
- 230000003044 adaptive effect Effects 0.000 claims description 41
- 238000001914 filtration Methods 0.000 claims description 36
- 238000012545 processing Methods 0.000 claims description 19
- 230000009191 jumping Effects 0.000 claims description 15
- 238000002474 experimental method Methods 0.000 claims description 9
- 238000004364 calculation method Methods 0.000 claims description 3
- 235000012434 pretzels Nutrition 0.000 claims description 3
- 238000011084 recovery Methods 0.000 abstract description 4
- 238000010586 diagram Methods 0.000 description 3
- 238000007781 pre-processing Methods 0.000 description 3
- 230000005540 biological transmission Effects 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 230000006978 adaptation Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000011109 contamination Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/70—Denoising; Smoothing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/12—Fingerprints or palmprints
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20024—Filtering details
- G06T2207/20032—Median filtering
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Human Computer Interaction (AREA)
- Multimedia (AREA)
- Collating Specific Patterns (AREA)
Abstract
The invention discloses an effective removing method for salt-and-pepper noise of fingerprint images, which comprises the following steps: 1) Setting an initial state value, 2) detecting whether all points in the fingerprint image are candidate noise points; 3) Detecting whether all candidate noise points in the fingerprint image are real noise points according to the relation between the minimum absolute brightness difference and the self-adaptive threshold value; 4) Selecting a denoising window by using the information of the extreme value pixel density in the window, and replacing the pixel value of the real noise point; 5) And sliding the filter window until the whole fingerprint image is processed, so that the salt and pepper noise of the fingerprint image is effectively removed. According to the characteristic of the lines on the fingerprint image, the invention can use the local detail information of the fingerprint image to solve the problems of removing salt and pepper noise and protecting details on the fingerprint image, and simultaneously promote the recovery effect of the fingerprint image when the fingerprint image is polluted by the salt and pepper noise, so that more accurate fingerprint characteristic identification can be realized.
Description
Technical Field
The invention relates to the technical field of fingerprint image processing, in particular to an effective removing method for salt-pepper noise of a fingerprint image.
Background
A complete fingerprint identification system comprises four steps of acquisition, preprocessing, feature extraction and feature matching. In these steps, the preprocessing of the fingerprint image is particularly important, and the effect of feature extraction and matching of the subsequent fingerprint image is directly affected. In preprocessing of fingerprint images, denoising of the images is very important. In the processes of fingerprint image acquisition, transmission, reception and processing, bit errors may occur due to factors such as uneven illumination or atmospheric interference, sensor noise, channel transmission errors and the like, so that impulse noise is introduced into the fingerprint image, the quality of the acquired image is reduced, and the fingerprint image is blurred. Salt and pepper noise can damage the image and thus noise reduction is an essential operation. In removing salt and pepper noise, there are four common methods: 1. all pixels in the image are directly processed, and non-noise points are also processed by the method to distort the image; 2. the points with maximum and minimum pixel values or points with pixel points of 0 and 255 in the processing window of the image are directly regarded as noise points, and some signal points with extreme pixel values are judged as noise by the method, so that the image is distorted; 3. the noise point is judged by comparing a preset fixed threshold with an index of an image processing window, the adaptability of the fixed threshold in the method to local details of the image is poor, and a proper threshold is difficult to find; 4. the noise point is judged by comparing the adaptive threshold with the index of the image processing window, but the effect is still required to be improved due to the influence of the problem of the adaptation of the local noise density of the image.
Disclosure of Invention
The invention aims to overcome the defects and shortcomings of the prior art, and provides an effective removing method for salt-pepper noise of a fingerprint image, which can improve the problems of removing the salt-pepper noise and protecting details on the fingerprint image by using the local detail information of the fingerprint image according to the characteristics of lines on the fingerprint image, and simultaneously improves the recovery effect of the fingerprint image when the fingerprint image is polluted by the salt-pepper noise, so that more accurate fingerprint characteristic identification can be further realized.
In order to achieve the above purpose, the technical scheme provided by the invention is as follows: the effective removing method for the salt-pepper noise of the fingerprint image comprises the following steps:
1) For a single fingerprint image, setting the width and the height of an initial filter window and setting the maximum width and the maximum height of the filter window;
2) Sliding a filtering window in the fingerprint image, judging whether the central point is a candidate noise point, if not, jumping to the step 5), and if so, continuing to execute the step 3);
3) Judging whether adjacent pixel points capable of calculating the minimum absolute brightness difference exist in the filter window, if the minimum absolute brightness difference is not smaller than the self-adaptive threshold, expanding the filter window, re-executing the step 3), if the minimum absolute brightness difference is larger than or equal to the self-adaptive threshold, replacing the central pixel with the average value of pixels in the 3×3 filter window with the current central pixel as the center, skipping the step 5), if the minimum absolute brightness difference is smaller than the self-adaptive threshold, then the candidate noise point is not the real noise point, skipping the step 5), if the minimum absolute brightness difference is larger than or equal to the self-adaptive threshold, then the candidate noise point is the real noise point, and continuing to execute the step 4);
4) For the real noise point, determining whether the filter window needs to be enlarged according to the comparison of the density of the extremum pixels in the filter window and the density threshold, if the density of the extremum pixels in the filter window is greater than or equal to the density threshold and the filter window does not reach the maximum value, enlarging the filter window and jumping to the step 3), if the density of the extremum pixels in the filter window is greater than or equal to the density threshold and the filter window reaches the maximum value, carrying out pixel replacement on the real noise point by using the median value of the non-extremum pixels in the filter window, continuing to execute the step 5), and if the density of the extremum pixels in the filter window is less than the density threshold, replacing the noise point by using the median value of the non-extremum pixels in the filter window, and continuing to execute the step 5);
5) If the fingerprint image is not processed, jumping back to the step 2), and if the fingerprint image is processed, outputting the fingerprint image, so that the effective removal of salt-pepper noise of the fingerprint image is realized.
Further, in step 1), the filter window refers to a rectangular window with equal width and height centered on the pixel point currently being processed, the initial size w=3 is set, and the maximum width and height W are set max ＝7。
Further, in step 2), if the center pixel value P (x, y) of the filter window currently being processed is located at the maximum pixel value P in the window max And a minimum pixel value P min If the center point is a non-noise point and jumps to step 5), if the filter window center pixel value P (x, y) =p currently being processed min Or P (x, y) =p max Then the center point is the candidate noise point and step 3) is continued.
Further, the step 3) includes the steps of:
3.1 If the pixel values of all points in the filter window are the maximum pixel value or the minimum pixel value at this time, and the processing window does not reach the maximum width and height W max =7, then the filtering is expandedThe window size is (w+2), step 3) is re-executed if the pixel values of all points in the filter window are the maximum pixel value or the minimum pixel value at this time, and the processing window has reached the maximum width and height W max =7, then replacing the center pixel with the average value of all pixels in the 3×3 filter window centered on the current center pixel, skipping to step 5); if the point with the non-maximum pixel value and the non-minimum pixel value exists in the filtering window at the moment, calculating the absolute value of the difference between the candidate noise point and the pixel values of other points with the non-maximum pixel value and the non-minimum pixel value in the filtering window, taking the minimum value as the minimum absolute brightness difference MABD value, wherein the magnitude of MABD represents the correlation between the candidate noise point and the adjacent non-noise pixels in the filtering window, and the smaller the MABD is, the higher the correlation between the center pixel and the adjacent non-noise pixels is, the larger the MABD is, and the lower the correlation between the center pixel and the adjacent non-noise pixels is;
3.2 The adaptive threshold T is not a fixed value, but introduces the concept of extreme pixel density, wherein the extreme pixel density refers to the proportion of points with pixel values of 0 and 255 in a filter window to the size of the filter window, the adaptive threshold T changes according to the extreme pixel density of the current window, and the local information in the image is fully utilized, so that the adaptive threshold T is adaptive to the local extreme pixel density in the image, and the adaptive threshold T is defined as a linear monotonically increasing function:
T＝d*k+b
where k and b are parameter values to be determined, d refers to the pixel density of the extremum of the window, and the setting of the adaptive threshold T takes into account: the smaller the extreme pixel density is, the more non-extreme pixel points exist around the candidate noise points, and the more non-extreme pixel points can be used for calculating the brightness difference, so that the possibility of the calculated MABD approaching to the real situation when the picture is not polluted by the salt and pepper noise is higher; the larger the extreme pixel density is, the fewer the non-extreme pixels exist around the candidate noise point, and the fewer the non-extreme pixels can be used to calculate the brightness difference, so the calculated MABD tends to be far away from the real situation when the picture is not contaminated by the pretzel noise, so the linear relation is adopted, in order to reduce the calculation amount, the parameters k and b need to be determined experimentally, the fingerprint image is tested to find the suitable parameters k and b, and in the experiment, the k and b are considered to respectively obtain the best image processing effect near a certain fixed value because: as k and b are increased, the adaptive threshold T is increased, more noise points are judged as signal points, and the false judgment probability is increased; as k and b decrease, the adaptive threshold T will decrease, more signal points will be judged as noise, and the misjudgment probability will increase, so k and b will reach the optimal processing effect when reaching a certain value, and k and b will cause the effect to be poor no matter becoming larger or smaller, the specific experimental steps are as follows:
the method comprises the steps of carrying out experiments on a single fingerprint image, and simultaneously obtaining an original fingerprint image without salt and pepper noise, wherein in order to better embody and compare a filtering effect, the effect processed by an algorithm is embodied by PSNR (peak signal to noise ratio), wherein the PSNR is used for measuring the approaching degree of a restored image and the original image, and the bigger the PSNR is, the closer the restored image and the original image are; PSNR is obtained from the following formula:
wherein:
where MSE represents the mean square error, mxn represents the product of the width and height of the image, f (i, j) represents the pixel value of the point of the fingerprint image without salt and pepper noise, g (i, j) represents the pixel value of the point of the single fingerprint image, i represents the number of rows of the image, and j represents the number of columns of the image;
comparing PSNR values obtained under different parameters k and b, taking the parameters k and b at the highest PSNR as parameter values of the experiment, obtaining an adaptive threshold T adaptive to local features of an image by a formula T=d x k+b of the adaptive threshold T, and jumping to the step 5) when MABD is smaller than the adaptive threshold T, wherein the candidate noise point is not a real noise point, and the pixel value is unchanged; when MABD is greater than or equal to the adaptive threshold T, the candidate noise point is a true noise point, and step 4 is continued.
Further, in step 4), for the true noise point obtained by step 3), a judgment is performed: if the density of the extreme pixel of the filtering window is less than 50% of the density threshold value, replacing the noise pixel with the median value of the non-extreme pixel of the filtering window, otherwise, if the density of the extreme pixel of the 3×3 window is greater than or equal to 50%, expanding the filtering window to be 5×5, and jumping to the step 3); if the pixel density of the extreme value of the filtering window is less than 75% of the density threshold value, namely the median value of the non-extreme value pixel of the current window is used for replacing the noise pixel, otherwise, if the pixel density of the extreme value of the 5 x 5 window is more than or equal to 75%, the maximum value of the filtering window is 7 x 7, and the step 3) is skipped; if the filter window is 7×7, the median value of the non-extremum pixels of the current window is used as output if the filter window extremum pixel density is less than 100% of the density threshold, that is, as long as the non-extremum pixels exist in the filter window, and if the filter window extremum pixel density is greater than or equal to 100% of the density threshold, that is, no non-extremum pixels exist in the filter window, the filter window is not expanded any more, the noise pixels are replaced by the average value of all pixels in the current 3×3 filter window, and the step 5 is continued.
Further, in step 5), checking whether all pixels of the current whole fingerprint image have been completely detected and filtered, and if not, jumping to step 2); and if the fingerprint image is completely finished, outputting the filtered fingerprint image, so that the effective removal of salt and pepper noise of the fingerprint image is realized.
Compared with the prior art, the invention has the following advantages and beneficial effects:
1. the invention adopts the self-adaptive threshold value according to the extreme pixel density change of the filter window for the first time, fully considers the local information in the fingerprint image, gives consideration to denoising and detail keeping, avoids the characteristic of poor adaptability of the fixed threshold value to the image, and has good processing effect on the local part with noise of different densities in the image.
2. The invention is designed aiming at the characteristics of the fingerprint image in removing the salt and pepper noise, and compared with other filtering methods for removing the salt and pepper noise, the quality of removing the salt and pepper noise on the fingerprint image is improved.
3. The invention has good recovery quality under the condition that the fingerprint image is polluted by low, medium and high density noise, so that the invention can be better matched with the existing fingerprint identification system algorithm, thereby realizing higher fingerprint matching efficiency.
4. The invention is suitable for fingerprint images with different qualities obtained by different fingerprint collectors, and has universality.
Drawings
FIG. 1 is a general flow chart of the method of the present invention.
FIG. 2 is a flowchart showing the steps of the method of the present invention.
Fig. 3 is a schematic diagram of a fingerprint image contaminated with 80% density noise.
Fig. 4 is a schematic diagram of a 3×3 filter window selected from fig. 3.
Fig. 5 is a schematic view of a 5 x 5 window selected from fig. 3 and extended from fig. 4.
Fig. 6 is a schematic view of a 7 x 7 window selected from fig. 3 and extended from fig. 5.
FIG. 7 is a recovery chart of FIG. 3 after denoising by the method of the present invention.
Fig. 8 is an original view of fig. 3, which is not contaminated with salt and pepper noise.
Detailed Description
The present invention will be described in further detail with reference to examples and drawings, but embodiments of the present invention are not limited thereto.
Referring to fig. 1 to 8, the present embodiment provides an effective removing method for salt-pepper noise of fingerprint images, which specifically includes the following steps:
1) For a single fingerprint image, setting the width and the height of an initial filter window to be W=3, and setting the maximum width and the maximum height to be W max =7. The filtering window refers to the current processingA rectangular window with equal width and height centered on the pixel point, the initial size is set to w=3, and the maximum width and height are set to W max =7. For example, fig. 3 is a fingerprint image contaminated by 80% salt and pepper noise, fig. 4 is a schematic diagram of a filtering window in fig. 3, and both pixel value 0 and pixel value 255 are caused by the contamination of salt and pepper noise, and the remaining points are points on the original fingerprint image.
2) Sliding a filter window in the fingerprint image, calculating the maximum pixel value P of all points in the filter window max And a minimum pixel value P min If the pixel value P (x, y) of the center point of the filter window is at the maximum pixel value P in the window max And a minimum pixel value P min If the filter window center point is a non-noise point and jumps to step 5), if the pixel value P (x, y) =p of the filter window center point min Or P (x, y) =p max Then, the center point of the filter window is the candidate noise point, and the step 3) is continued.
For example, as shown in fig. 4, the maximum pixel value P in the filter window max =255, minimum pixel value P min =0, then this pixel is a candidate noise point, and step 3) is continued.
3) Judging whether adjacent pixel points which can be used for calculating the minimum absolute brightness difference exist in the filtering window, if the minimum absolute brightness difference is not smaller than the self-adaptive threshold, expanding the filtering window, re-executing the step 3), if the minimum absolute brightness difference is larger than or equal to the self-adaptive threshold, replacing the central pixel with the average value of pixels in the 3×3 filtering window with the current central pixel as the center, skipping the step 5), if the minimum absolute brightness difference is smaller than the self-adaptive threshold, then the candidate noise point is not the real noise point, and if the minimum absolute brightness difference is larger than or equal to the self-adaptive threshold, then the step 4) is continued to be executed. Specifically, the method comprises the following steps:
3.1 If the pixel values of all points in the filter window are the maximum pixel value orMinimum pixel value and processing window does not reach maximum W max If the pixel values of all points in the filter window are the maximum pixel value or the minimum pixel value, and the processing window has reached the maximum W max =7, then replacing the center pixel with the average value of all pixels in the 3×3 filter window centered on the current center pixel, skipping to step 5); if the point with the non-maximum pixel value and the non-minimum pixel value exists in the filtering window at this time, the absolute value of the difference between the candidate noise point and the pixel values of other points with the non-maximum pixel value and the non-minimum pixel value in the filtering window is calculated, the minimum value is taken as the minimum absolute brightness difference MABD value, the magnitude of MABD represents the correlation between the candidate noise point and the adjacent non-noise pixels in the filtering window, the smaller the MABD is, the higher the correlation between the center pixel and the adjacent non-noise pixels is, the larger the MABD is, and the correlation between the center pixel and the adjacent non-noise pixels is lower. For example, as shown in fig. 4, when there is a point of non-maximum pixel value and non-minimum pixel value in the filtering window, the calculated mabd=min { |28-0| } =28.
3.2 The adaptive threshold T is not a fixed value, but introduces the concept of extreme pixel density, wherein the extreme pixel density refers to the proportion of points with pixel values of 0 and 255 in a filter window to the size of the filter window, the adaptive threshold T changes according to the extreme pixel density of the current window, and the local information in the image is fully utilized, so that the adaptive threshold T is adaptive to the local extreme pixel density in the image, and the adaptive threshold T is defined as a linear monotonically increasing function:
T＝d*k+b
where k and b are parameter values to be determined, d refers to the pixel density of the extremum of the window, and the setting of the adaptive threshold T takes into account: the smaller the extreme pixel density is, the more non-extreme pixel points exist around the candidate noise points, and the more non-extreme pixel points can be used for calculating the brightness difference, so that the possibility of the calculated MABD approaching to the real situation when the picture is not polluted by the salt and pepper noise is higher; the larger the extreme pixel density is, the fewer the non-extreme pixels exist around the candidate noise point, and the fewer the non-extreme pixels are available for calculating the brightness difference, so the calculated MABD tends to be far away from the real situation when the picture is not contaminated by the pretzel noise, so the linear relation is adopted, in order to reduce the calculation amount, the parameters k and b need to be determined experimentally, the fingerprint image is tested to find the suitable parameters k and b, and in the experiment, the k and b are considered to respectively obtain the best image processing effect near a certain fixed value because: as k and b are increased, the adaptive threshold T is increased, more noise points are judged as signal points, and the false judgment probability is increased; as k and b decrease, the adaptive threshold T will decrease, more signal points will be judged as noise, and the misjudgment probability will increase, so k and b will reach the optimal processing effect when reaching a certain value, and k and b will cause the effect to be poor no matter becoming larger or smaller, the specific experimental steps are as follows:
and (3) carrying out experiments on the single fingerprint image, and simultaneously acquiring an original fingerprint image without salt and pepper noise, wherein in order to better embody and compare the filtering effect, the effect processed by the algorithm is embodied by PSNR (peak signal to noise ratio), wherein the PSNR is used for measuring the approaching degree of the restored image and the original image, and the bigger the PSNR, the closer the restored image and the original image are. PSNR is obtained from the following formula:
wherein:
where MSE denotes the mean square error, mxn denotes the product of the width and height of the image, f (i, j) denotes the pixel value of the point of the fingerprint image without salt and pepper noise, g (i, j) denotes the pixel value of the point of the single fingerprint image, where i denotes the number of rows of the image, and j denotes the number of columns of the image.
Comparing PSNR values obtained through processing under different parameters k and b, taking the parameters k and b at the highest PSNR as parameter values of the experiment, obtaining an adaptive threshold T adaptive to local features of an image by a formula T=d=k+b of the adaptive threshold T, and jumping to the step 5) when the MABD is smaller than the adaptive threshold T, wherein the candidate noise point is not a real noise point, and the pixel value is kept unchanged; when the MABD is equal to or greater than the adaptive threshold T, the candidate noise point is a true noise point, and step 4 is continued.
For example, as shown in fig. 4, a single fingerprint image is shown, and fig. 8 is a fingerprint image without salt and pepper noise, and for determining parameters k and b, a control variable method is adopted, and a mode of fixing one parameter and changing the other parameter is adopted to perform a large-scale parameter determination. First, with a large step size of 5, the parameter k starts at 5 and increases from 5 to 20, and b starts at 5 and increases from 5 to 25. The result is a maximum PSNR value that achieves a maximum psnr= 28.5034 at k=5, b=20. The parameter control test by the controlled variable method using 1 as a step is continued around the k and b values, and k ranges from 1 to 9,b from 16 to 24, and as a result, the maximum PSNR value of psnr= 28.5034 is obtained when k=5 and b=20, and therefore, the parameter k=5 and b=20 is calculated to be t=5× (8/9) +20=24.44 and mabd=28 > T for fig. 3, and the calculated adaptive threshold T is calculated to be t=5× (8/9) +20=24.44 and mabd=28 > T for fig. 4, and the candidate noise point is determined to be the true noise point.
4) For the real noise point, determining whether the filter window needs to be enlarged according to the comparison of the density of the extremum pixels in the filter window and the density threshold, if the density of the extremum pixels in the filter window is greater than or equal to the density threshold and the filter window does not reach the maximum value, enlarging the filter window and jumping to the step 3), if the density of the extremum pixels in the filter window is greater than or equal to the density threshold and the filter window reaches the maximum value, carrying out pixel replacement on the real noise point by using the median value of the non-extremum pixels in the filter window, continuing to execute the step 5), and if the density of the extremum pixels in the filter window is less than the density threshold, replacing the noise point by using the median value of the non-extremum pixels in the filter window, and continuing to execute the step 5);
for the true noise point obtained by step 3), the determination is performed: if the pixel density is equal to or higher than 50%, and if the pixel density is equal to or higher than 50%, the pixel density is equal to or higher than 5%, and the pixel density is equal to or higher than 50%, and the pixel density is equal to or higher than 5%, and the process jumps to step 3); if the pixel density of the extreme value pixel of the filtering window is less than 75% of the density threshold value, namely replacing the noise pixel with the median value of the non-extreme value pixel of the current window, otherwise, if the pixel density of the extreme value pixel of the 5 x 5 window is more than or equal to 75%, expanding the window to be 7 x 7 which is the set maximum value, and jumping to the step 3); if the filter window is 7×7, the median value of the non-extremum pixels of the current window is used as output if the filter window extremum pixel density is less than 100% of the density threshold, that is, as long as the non-extremum pixels exist in the filter window, and if the filter window extremum pixel density is greater than or equal to 100% of the density threshold, that is, no non-extremum pixels exist in the filter window, the window is not expanded any more, the noise pixels are replaced by the average value of all pixels in the current 3×3 filter window, and the step 5 is continued.
For example, in fig. 4, the extremum density of the filtering window is greater than 50%, the expansion window is 5×5, as shown in fig. 5, the step 3) is skipped, the real noise point is still determined after the step 3), the extremum density of the filtering window is greater than 75%, the expansion window is 7×7, as shown in fig. 6, the step 3) is skipped, the real noise point is still determined after the step 3), the non-extremum pixel exists in the current window, and the noise point is replaced by the median value of the non-extremum pixel of the current window, namely 29.
5) Checking whether all pixel points of the current whole fingerprint image are completely detected and filtered, if not, jumping back to the step 2), and if the fingerprint image is completely processed, outputting the filtered fingerprint image, so as to effectively remove salt and pepper noise of the fingerprint image. For example, in the fingerprint image of fig. 3, the filtering window is continuously slid to the next point to be processed in the image, the method is circularly executed until the whole image is processed, and the obtained result is shown in fig. 7, so that the effective removal of salt and pepper noise of the fingerprint image is realized.
The above examples are preferred embodiments of the present invention, but the embodiments of the present invention are not limited to the above examples, and any other changes, modifications, substitutions, combinations, and simplifications that do not depart from the spirit and principle of the present invention should be made in the equivalent manner, and the embodiments are included in the protection scope of the present invention.
Claims (4)
1. The effective removing method for the salt and pepper noise of the fingerprint image is characterized by comprising the following steps of:
1) For a single fingerprint image, setting the width and the height of an initial filter window and setting the maximum width and the maximum height of the filter window;
2) Sliding a filtering window in the fingerprint image, judging whether the central point is a candidate noise point, if not, jumping to the step 5), and if so, continuing to execute the step 3);
3) Judging whether adjacent pixel points capable of calculating the minimum absolute brightness difference exist in the filter window, if the minimum absolute brightness difference is not smaller than the self-adaptive threshold, expanding the filter window, re-executing the step 3), if the minimum absolute brightness difference is larger than or equal to the self-adaptive threshold, replacing the central pixel with the average value of pixels in the 3×3 filter window with the current central pixel as the center, skipping the step 5), if the minimum absolute brightness difference is smaller than the self-adaptive threshold, then the candidate noise point is not the real noise point, skipping the step 5), if the minimum absolute brightness difference is larger than or equal to the self-adaptive threshold, then the candidate noise point is the real noise point, and continuing to execute the step 4); the method comprises the following steps:
3.1 If the pixel values of all points in the filter window are the maximum pixel value or the minimum pixel value at this time, and the processing window does not reach the maximum width and height W max =7, then the size of the filter window is enlarged to w+2, step 3) is re-executed, if all points in the filter window are nowIs the maximum pixel value or the minimum pixel value, and the processing window has reached the maximum width and height W max =7, then replacing the center pixel with the average value of all pixels in the 3×3 filter window centered on the current center pixel, skipping to step 5); if the point with the non-maximum pixel value and the non-minimum pixel value exists in the filtering window at the moment, calculating the absolute value of the difference between the candidate noise point and the pixel values of other points with the non-maximum pixel value and the non-minimum pixel value in the filtering window, taking the minimum value as the minimum absolute brightness difference MABD value, wherein the magnitude of MABD represents the correlation between the candidate noise point and the adjacent non-noise pixels in the filtering window, and the smaller the MABD is, the higher the correlation between the center pixel and the adjacent non-noise pixels is, the larger the MABD is, and the lower the correlation between the center pixel and the adjacent non-noise pixels is;
3.2 The adaptive threshold T is not a fixed value, but introduces the concept of extreme pixel density, wherein the extreme pixel density refers to the proportion of points with pixel values of 0 and 255 in a filter window to the size of the filter window, the adaptive threshold T changes according to the extreme pixel density of the current window, and the local information in the image is fully utilized, so that the adaptive threshold T is adaptive to the local extreme pixel density in the image, and the adaptive threshold T is defined as a linear monotonically increasing function:
T＝d*k+b
where k and b are parameter values to be determined, d refers to the pixel density of the extremum of the window, and the setting of the adaptive threshold T takes into account: the smaller the extreme pixel density is, the more non-extreme pixel points exist around the candidate noise points, and the more non-extreme pixel points can be used for calculating the brightness difference, so that the possibility of the calculated MABD approaching to the real situation when the picture is not polluted by the salt and pepper noise is higher; the larger the extreme pixel density is, the fewer the non-extreme pixels exist around the candidate noise point, and the fewer the non-extreme pixels can be used to calculate the brightness difference, so the calculated MABD tends to be far away from the real situation when the picture is not contaminated by the pretzel noise, so the linear relation is adopted, in order to reduce the calculation amount, the parameters k and b need to be determined experimentally, the fingerprint image is tested to find the suitable parameters k and b, and in the experiment, the k and b are considered to respectively obtain the best image processing effect near a certain fixed value because: as k and b are increased, the adaptive threshold T is increased, more noise points are judged as signal points, and the false judgment probability is increased; as k and b decrease, the adaptive threshold T will decrease, more signal points will be judged as noise, and the misjudgment probability will increase, so k and b will reach the optimal processing effect when reaching a certain value, and k and b will cause the effect to be poor no matter becoming larger or smaller, the specific experimental steps are as follows:
the method comprises the steps of carrying out experiments on a single fingerprint image, and simultaneously obtaining an original fingerprint image without salt and pepper noise, wherein in order to better embody and compare a filtering effect, the effect processed by an algorithm is embodied by PSNR (peak signal to noise ratio), wherein the PSNR is used for measuring the approaching degree of a restored image and the original image, and the bigger the PSNR is, the closer the restored image and the original image are; PSNR is obtained from the following formula:
wherein:
where MSE represents the mean square error, mxn represents the product of the width and height of the image, f (i, j) represents the pixel value of the point of the fingerprint image without salt and pepper noise, g (i, j) represents the pixel value of the point of the single fingerprint image, i represents the number of rows of the image, and j represents the number of columns of the image;
comparing PSNR values obtained under different parameters k and b, taking the parameters k and b at the highest PSNR as parameter values of the experiment, obtaining an adaptive threshold T adaptive to local features of an image by a formula T=d x k+b of the adaptive threshold T, and jumping to the step 5) when MABD is smaller than the adaptive threshold T, wherein the candidate noise point is not a real noise point, and the pixel value is unchanged; when the MABD is greater than or equal to the adaptive threshold T, the candidate noise point is a real noise point, and the step 4) is continuously executed;
4) For the real noise point, determining whether the filter window needs to be enlarged according to the comparison of the density of the extremum pixels in the filter window and the density threshold, if the density of the extremum pixels in the filter window is greater than or equal to the density threshold and the filter window does not reach the maximum value, enlarging the filter window and jumping to the step 3), if the density of the extremum pixels in the filter window is greater than or equal to the density threshold and the filter window reaches the maximum value, carrying out pixel replacement on the real noise point by using the median value of the non-extremum pixels in the filter window, continuing to execute the step 5), and if the density of the extremum pixels in the filter window is less than the density threshold, replacing the noise point by using the median value of the non-extremum pixels in the filter window, and continuing to execute the step 5);
for the true noise point obtained by step 3), the determination is performed: if the density of the extreme pixel of the filtering window is less than 50% of the density threshold value, replacing the noise pixel with the median value of the non-extreme pixel of the filtering window, otherwise, if the density of the extreme pixel of the 3×3 window is greater than or equal to 50%, expanding the filtering window to be 5×5, and jumping to the step 3); if the pixel density of the extreme value of the filtering window is less than 75% of the density threshold value, namely the median value of the non-extreme value pixel of the current window is used for replacing the noise pixel, otherwise, if the pixel density of the extreme value of the 5 x 5 window is more than or equal to 75%, the maximum value of the filtering window is 7 x 7, and the step 3) is skipped; if the current filter window is 7×7, if the density of the extremum pixels of the filter window is less than 100% of the density threshold, that is, if there is no extremum pixel in the filter window, the median value of the extremum pixels of the current window is used as output, if the density of the extremum pixels of the filter window is greater than or equal to 100% of the density threshold, that is, if there is no extremum pixel in the filter window, the filter window is not enlarged any more, the average value of all pixels in the current 3×3 filter window is used to replace the noise pixel, and the step 5) is continued;
5) If the fingerprint image is not processed, jumping back to the step 2), and if the fingerprint image is processed, outputting the fingerprint image, so that the effective removal of salt-pepper noise of the fingerprint image is realized.
2. The method as claimed in claim 1, wherein in the step 1), the filter window is a rectangular window with equal width and height centered on the pixel point currently processed, the initial size w=3 is set, and the maximum width and height W are set max ＝7。
3. The method for efficient removal of salt and pepper noise as claimed in claim 2, characterized in that in step 2) if the center pixel value P (x, y) of the filter window currently being processed is located at the maximum pixel value P in the window max And a minimum pixel value P min If the center point is a non-noise point and jumps to step 5), if the filter window center pixel value P (x, y) =p currently being processed min Or P (x, y) =p max Then the center point is the candidate noise point and step 3) is continued.
4. The method for efficient removal of salt and pepper noise in fingerprint images according to claim 1, characterized in that in step 5) it is checked whether all pixels of the current whole fingerprint image have been completely detected and filtered, and if not, it goes to step 2); and if the fingerprint image is completely finished, outputting the filtered fingerprint image, so that the effective removal of salt and pepper noise of the fingerprint image is realized.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210501241.9A CN115063302B (en) | 2022-05-10 | 2022-05-10 | Effective removing method for spiced salt noise of fingerprint image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210501241.9A CN115063302B (en) | 2022-05-10 | 2022-05-10 | Effective removing method for spiced salt noise of fingerprint image |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115063302A CN115063302A (en) | 2022-09-16 |
| CN115063302B true CN115063302B (en) | 2024-03-29 |
Family
ID=83199229
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210501241.9A Active CN115063302B (en) | 2022-05-10 | 2022-05-10 | Effective removing method for spiced salt noise of fingerprint image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115063302B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115829873B (en) * | 2022-12-13 | 2023-12-19 | 深圳市宏电技术股份有限公司 | Image restoration method and processing system |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101094312A (en) * | 2006-06-20 | 2007-12-26 | 西北工业大学 | Self-adapting method for filtering image with edge being retained |
| CN103871034A (en) * | 2014-03-22 | 2014-06-18 | 四川大学 | Self-adapting filtering method for salt and pepper noise of image |
| CN105243649A (en) * | 2015-11-09 | 2016-01-13 | 天津大学 | Image denoising method based on secondary noise point detection |
| CN109961411A (en) * | 2019-03-12 | 2019-07-02 | 浙江工业大学之江学院 | Non-lower sampling shears wave conversion medicine CT image denoising method |
| CN113989168A (en) * | 2021-11-02 | 2022-01-28 | 华北电力大学（保定） | Self-adaptive non-local mean filtering method for salt and pepper noise |
-
2022
- 2022-05-10 CN CN202210501241.9A patent/CN115063302B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101094312A (en) * | 2006-06-20 | 2007-12-26 | 西北工业大学 | Self-adapting method for filtering image with edge being retained |
| CN103871034A (en) * | 2014-03-22 | 2014-06-18 | 四川大学 | Self-adapting filtering method for salt and pepper noise of image |
| CN105243649A (en) * | 2015-11-09 | 2016-01-13 | 天津大学 | Image denoising method based on secondary noise point detection |
| CN109961411A (en) * | 2019-03-12 | 2019-07-02 | 浙江工业大学之江学院 | Non-lower sampling shears wave conversion medicine CT image denoising method |
| CN113989168A (en) * | 2021-11-02 | 2022-01-28 | 华北电力大学（保定） | Self-adaptive non-local mean filtering method for salt and pepper noise |
Non-Patent Citations (1)
| Title |
|---|
| 一种有效去除椒盐噪声的开关中值滤波器;许光宇;;安徽理工大学学报(自然科学版);20170115(01);第37-43页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115063302A (en) | 2022-09-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110796615B (en) | Image denoising method, device and storage medium | |
| CN105243649A (en) | Image denoising method based on secondary noise point detection | |
| CN102479383B (en) | Method and device for removing salt and pepper noise | |
| CN115063302B (en) | Effective removing method for spiced salt noise of fingerprint image | |
| US20110164284A1 (en) | Image processing apparatus and method | |
| CN112862753B (en) | Noise intensity estimation method and device and electronic equipment | |
| Raza et al. | High density salt and pepper noise removal through decision based partial trimmed global mean filter | |
| CN106530254A (en) | Algorithm for inhibiting mixed noise of images based on wavelet threshold function and improved median filtering fusion | |
| CN111612773B (en) | Thermal infrared imager and real-time automatic blind pixel detection processing method | |
| US20090316049A1 (en) | Image processing apparatus, image processing method and program | |
| JP2015005959A (en) | Image processing method, apparatus, and program | |
| CN112215764A (en) | Image processing method based on median filtering improved algorithm | |
| US8165421B2 (en) | Method and apparatus for image processing by using stored image | |
| CN111242884B (en) | Image dead pixel detection and correction method and device, storage medium and camera equipment | |
| CN115829873B (en) | Image restoration method and processing system | |
| EP2226761B1 (en) | Methods and systems for filtering a digital signal | |
| US20070070428A1 (en) | Device and method for sharpening image signal | |
| Zhang et al. | Varying weight trimmed mean filter for the restoration of impulse noise corrupted images | |
| EP1654881A1 (en) | Block artifacts detection | |
| Marcel et al. | An efficient algorithm for removal of impulse noise using Adaptive Fuzzy Switching Weighted Median Filter | |
| CN112785513B (en) | Self-adaptive median filtering method for filtering impulse noise | |
| CN111223074B (en) | Method for detecting top concrete defect of concrete column member | |
| RU2405200C2 (en) | Method and device for fast noise filtration in digital images | |
| CN110533605B (en) | Accurate noise point calibration method | |
| KR20180085098A (en) | Image error mitigation method using neighboring pixels in noise environment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |